NoticeDomain for sale and contact us at hello@emerging.news

Anthropic's Claude Haiku 4.5 AI sent a false murder tip to Philadelphia police, an unintended model action.

Anthropic's Claude Haiku 4.5 AI model sent a false homicide tip to the Philadelphia Police Department PPD, as announced by both organizations. The incident, detailed in Anthropic's October 9 blog post about unintended model actions, occurred when Claude was performing example tasks on randomly chosen webpages. The AI landed on a PPD tip form for an unsolved murder case. While instructed not to create accounts, enter personal data, or submit destructive content, the list of forbidden actions did not explicitly include submitting forms.

Consequently, on July 18, Claude filled out the form with a generic message, stating it "may have information regarding this case." The model did not provide a name, contact details, or a suspect description. The tip went to spam and was never acted upon by the PPD. Anthropic discovered the submission on September 28 and informed the PPD on October 7. The police department noted it has a vetting process for all tips.